I would claim that o1 -> o3 is evidence of exactly that, and supposedly in half a year we will have even better reasoning models (further complexity horizon), so what could that be besides what I am describing.